When Generative Artificial Intelligence Makes Mistakes: A Hybrid Approach to Service Recovery From “Me” and “You” Perspectives
Dong Lv, Rui Sun, Qiuhua Zhu, Yue Cheng, Shukun Qin · Journal of Consumer Behaviour · 2025
ABSTRACT Generative artificial intelligence (GenAI) has demonstrated immense application value across various domains owing to its high‐quality content generation capabilities. However, avoiding service failures, such as AI hallucinations, remains challenging. Grounded in the dual‐process and cognitive appraisal theories of emotions, this study uses event‐related potential (ERP) experiments and scenario‐based questionnaires as part of a mixed‐method approach to investigate the impact of service recovery perspectives (first‐person vs. second‐person) on users' willingness to forgive. It investigates the moderating effect of service failure severity (minor vs. severe) and the mediating influence of perceived relief. The questionnaire findings reveal that the first‐person perspective in service recovery significantly enhances users' perceived relief, thereby increasing their willingness to forgive, especially for severe service failures. The ERP results indicate that first‐person service recovery induces larger P2 amplitudes and late positive potential components. This study contributes to the literature on service recovery perspectives and presents a novel research design that integrates subjective observations with objective physiological markers. It provides empirical evidence for designing more effective service recovery strategies for GenAI service providers, promoting the harmonious development of human–computer interactions.